IDEAS

Here are some ideas (en vrac) that makes pml2 what it is.

- Sub-typing without cohercion

- Pure-terms as objects of the logic

- Termination checker integrated to the system (also at the level of
  the model)

- inductive and coinductive and mixing both supported in a uniform way
  (termination checks the wellfoundedness of proofs)

- extensible records

- polymorphic variants

- reasonning on equivalence of programs with the restriction
  connectives in types

- membership connectives (aka singleton types) to encode dependent
  types

- call-by-value (and its realizability model)

- classical logic with lambda-mu (like call-cc) (pb with the above
  solved by the property below)

- [A]^^ inter V = [A] as main property of the model thanks to delta

- proof in pml looks nice

- higher or types like o => o inhabited my "macros"

- programs as proofs paradigm as opposed to proof as programs

- proofs are programs with all leafs unreachable ... so proof are dead
  code dans call to lemma in program can be removed at compile time

- we face undecidability, but only subtyping may loop => error messages
  as usual and fixed when PML loop

- surcharge et sous-typage ensemble (truc qui ne marchait pas dans PML1)